Discussion of heatsink designs

Sin22 · 09-11-2005, 09:11 AM

Thought this would be an interesting topic to bring back up.

In my experience, I've seen several heatsinks that boast impressive thermal resistance figures and in actual use also tend to have better load temperatures than a comparitive high-end heatsink.

However, when it boiled down to overclocking, sometimes the opponent would win. Case in point, the old Thermalright SLK900A vs Swiftech MCX462-V.

Well, recently a similar phenomenon has occured. Comparing a Coolermaster Hyper6 versus a Tuniq Tower120 HSF, load temp difference was almost 10degC in favor of the Tuniq.

Yet, currently, the Tower120 is unable to keep my Venice at 2.8GHz as what my old Hyper6 was capable of doing.

Now what I'm hoping to get going is a discussion on what sorts of designs, at least for the current heatpipe style coolers, work best in both promoting low temps, but more importantly, high overclocking headroom.

I'd like to get my hands on a Titan Vannessa L-Type HSF to see what its like. Thus, far I've only had a chance to try out XP90 or tower style heatpipe heatsinks and with varying degrees of overclocking headroom.

The L-Type uses a big ass heatpipe right over the core and this seems to gel with the theory of getting both low temps and high overclocking headroom due to the large amount of heat capable of being removed.

Your thoughts?

jaydee · 09-11-2005, 06:16 PM

I would like to know why a cooler that keeps the CPU colder doesn't overclock better. What would cause that... That is the question that needs to be answered before designs are considered. Obviously the design would have to be around that solution.

Sin22 · 09-11-2005, 06:31 PM

Summarised what I was trying to say very succinctly jaydee.

My initial thoughts had to do with heatspots/spikes on the die and how inidividual heatsinks coped with it. i.e. those with thicker bases managed to negate to some degree the effects of heatspots in comparison to those with thinner bps much like waterblocks (just as how the Cascade at some instances was shown to overclock better than comparitive blocks though temps were almost similar).

Long Haired Git · 09-11-2005, 07:45 PM

LRR and Cascade did well by focusing cooling to the CPU die, a big step from the "spiral" blocks before it.
Perhaps by then focusing the cooling more to the location of the temperatures sensors within the die a "better" heatsink/waterblock can be made....

Sin22 · 09-11-2005, 08:53 PM

The way I saw it is that the Cascade/Storm focused the cooling across the die effectively.

The same can be said for thicker BP waterblocks where the heat-distribution is more even like the MCW500x/MCW600x series.

However. in a sense, you can't really focus the cooling of a heatsink onto the location of the temperature sensors and what not, much like we could with waterblocks. (Cutouts in the base would be the exception much like the Cooljag CJC6166A).

That is what I felt was happening with the SLK900A & MCX462-V whereby the thicker baseplate of the MCX462-V allowed for better heat distribution and evening out of the heatspots.

Now though, with the heatpipe/tower heatsinks how would this be different with the added medium of a heatpipe?

To me, I've seen heatpipes as more like highways, just moving the heat from one point to another, allowing for greater cooling efficiency at the new location (and also expanding some heat through the actual movement). However, comparing the two heatsinks currently in question, the baseplates are almost identical in thickness, except the Hyper6 includes an aluminum heatsink above the heatpipe soldered area.

Would this addition in turn act as an increased thickness baseplate? For the Tuniq Tower120, the thickness of the base between the IHS and the bottom of the heatpipes appears to only be 1.5mm.

The bad thing in a sense with all this is that thermal resistance figures can be easily found and in sense double-checked versus other testbeds (though the die size is an issue), however, double-checking the overclocking effectiveness of a heatsink is quite based on anecdotal evidence.

Spot · 09-11-2005, 11:50 PM

My guess is.........

That explains the heat sink at the bottom of the hyper6. The extra metal below helps spread the heat load while the gas/liqiud in heat pipes takes their own sweet time pushing the heat through.

Looks like the copper base of the Tuniq has a smaller threshold on the amount of heat it can take down there. Therefore, a simple heatsink attached to the copper base at the bottom may just help.

Sin22 · 09-12-2005, 07:57 PM

Sound plausible. So the only thing to do is to solder/attach a heatsink to the base in some fashion to see if overall overclocking increases.

bobo5195 · 06-15-2007, 01:26 PM

I remember me and bill used to yabber on about how hot spots on a die are a problem.

I know that intel showed some pics of huge variations over a p4 die.

In order to measure the effect you would need to measure the less conductive area over the CPU die (which is unknown) or a heat plot of the base of the device. You could in theory match a heat plot to a CPU heat plot to see what cooler is best as thermal paste should easentially be a 1D conductor is applied right.

How to measure heat effects like this is a hard problem for someone with hobbist grade equipment. I think that you could develop an approach with different heat distributions over various heat dies to try and approximate the effect and develop "a number" to take this into account.

Heat pipes have various effects along their length and production run application of conductive elements presents problems. I have a crazy idea brewing that heat pipes might effect performance by lowering the fourier time (time taken for heat to reach one point from another) which will effect small scale changes.

BGP Spook · 06-19-2007, 12:58 AM

Quote:

Originally Posted by bobo5195

I have a crazy idea brewing that heat pipes might effect performance by lowering the fourier time (time taken for heat to reach one point from another) which will effect small scale changes.

I am wondering how you would justify the statement regarding heatpipes lowering the fourier time? A better question might be, what points are you referring to when you say "one point form another?" From the base to the cooling fins? I already know(or think I know) the principals behind and functioning of heatpipes.

I am wondering because I haven't taken thermo yet, only Uni. level Chem. and Phys. 1 and 2.

I did some quick searches and didn't come up with much explanation for "fourier time" other than it is named after Joseph Fourier and that there is a linear transform used to deconstruct periodic functions and a series named after Fourier as well.

bobo5195 · 06-19-2007, 12:46 PM

It was a bit of a bullpoop argument thinking on it but this is what i meant....

It a heat transfer effect not a maths weirdity.
I never really heard of the concept of fourier time outside of my lecture notes as well, so it might be called something else. My lecturer was a don of heat transfer but he used weird words.

Fourier time is the amount of time it takes heat to reach one location from another. E.g. i warm a bar one end how long does it take to reach the other. Applied to steady state theory if i apply a pulse (like a hot bit of processor warming up) whats the mean time for me to see an output of that pulse at the end. Its going to be governed by the bodies thermal mass and its conductivity.

In the case of a solid bit of copper its a diffusive process wasily workable (if like abit of maths and simplifying assumptions) from basic laws. In heat pipes you are dealing with a much slower mass transfer process.

Its introducing dynamic process into the mix (just to make the question harder). I'm unsure that it is really the case as the heat is going to quite rapidly diffuse. A little thought experiment proves that it is quite a weird problem that isn't really going to come up alot.

billbartuska · 06-19-2007, 01:23 PM

Quote:

Originally Posted by bobo5195

Ithe amount of time it takes heat to reach one location from another.

The Dice/LN2 guys that design the pots that hold the Dry Ice/LN2 spend alot of time getting mass/surface areas matched to the metal (Cu/Al) and the temps they are shooting for.
Mass to act as a cold sink and surface area to dissipate the heat into the cooling medium.

wildfrogman1 · 12-22-2007, 02:50 AM

http://www.blairwing.com/images/ULTCo2/9.jpg
I found this while looking around online. This seems relevant to the discussion. From what I've seen, and read certain heatpipe heatsink bases are overly thin. Sometimes to the point of base warping under high mounting pressure. So you need a thick enough base to keep from warping, yet focused cooling all across the die. The hottest die spots are not in the center of most cpu dies, but off on the side halfway between center and the corners.

I got a birthday notice in my email, and remembered I had an account here

Sooo I'll be more active.

koolguy · 11-11-2010, 02:10 PM

Hello everybody,
I want to start playing around with oil cooling. Now, Submerging a MB in oil is totally nuts...But..

What I am thinking is something like a water cooling system but running oil in it.

Does anybody know about any website to start reading about it?

Thank you in advance.

09-11-2005, 09:11 AM	#1
Sin22 Cooling Neophyte Join Date: May 2003 Location: Singapore Posts: 66	Discussion of heatsink designs Thought this would be an interesting topic to bring back up. In my experience, I've seen several heatsinks that boast impressive thermal resistance figures and in actual use also tend to have better load temperatures than a comparitive high-end heatsink. However, when it boiled down to overclocking, sometimes the opponent would win. Case in point, the old Thermalright SLK900A vs Swiftech MCX462-V. Well, recently a similar phenomenon has occured. Comparing a Coolermaster Hyper6 versus a Tuniq Tower120 HSF, load temp difference was almost 10degC in favor of the Tuniq. Yet, currently, the Tower120 is unable to keep my Venice at 2.8GHz as what my old Hyper6 was capable of doing. Now what I'm hoping to get going is a discussion on what sorts of designs, at least for the current heatpipe style coolers, work best in both promoting low temps, but more importantly, high overclocking headroom. I'd like to get my hands on a Titan Vannessa L-Type HSF to see what its like. Thus, far I've only had a chance to try out XP90 or tower style heatpipe heatsinks and with varying degrees of overclocking headroom. The L-Type uses a big ass heatpipe right over the core and this seems to gel with the theory of getting both low temps and high overclocking headroom due to the large amount of heat capable of being removed. Your thoughts?

09-11-2005, 07:45 PM	#4
Long Haired Git Cooling Savant Join Date: Jan 2003 Location: Sydney, Oz Posts: 336	LRR and Cascade did well by focusing cooling to the CPU die, a big step from the "spiral" blocks before it. Perhaps by then focusing the cooling more to the location of the temperatures sensors within the die a "better" heatsink/waterblock can be made.... __________________ Long Haired Git "Securing an environment of Windows platforms from abuse - external or internal - is akin to trying to install sprinklers in a fireworks factory where smoking on the job is permitted." (Prof. Gene Spafford) My Rig, in all its glory, can be seen best here AMD XP1600 @ 1530 Mhz \| Soyo Dragon + \| 256 Mb PC2700 DDRAM \| 2 x 40 Gb 7200rpm in Raid-0 \| Maze 2, eheim 1250, dual heater cores! \| Full specifications (PCDB)

06-15-2007, 01:26 PM	#8
bobo5195 Cooling Savant Join Date: Aug 2005 Location: uk Posts: 400	Re: Discussion of heatsink designs I remember me and bill used to yabber on about how hot spots on a die are a problem. I know that intel showed some pics of huge variations over a p4 die. In order to measure the effect you would need to measure the less conductive area over the CPU die (which is unknown) or a heat plot of the base of the device. You could in theory match a heat plot to a CPU heat plot to see what cooler is best as thermal paste should easentially be a 1D conductor is applied right. How to measure heat effects like this is a hard problem for someone with hobbist grade equipment. I think that you could develop an approach with different heat distributions over various heat dies to try and approximate the effect and develop "a number" to take this into account. Heat pipes have various effects along their length and production run application of conductive elements presents problems. I have a crazy idea brewing that heat pipes might effect performance by lowering the fourier time (time taken for heat to reach one point from another) which will effect small scale changes.

06-19-2007, 12:46 PM	#10
bobo5195 Cooling Savant Join Date: Aug 2005 Location: uk Posts: 400	Re: Discussion of heatsink designs It was a bit of a bullpoop argument thinking on it but this is what i meant.... It a heat transfer effect not a maths weirdity. I never really heard of the concept of fourier time outside of my lecture notes as well, so it might be called something else. My lecturer was a don of heat transfer but he used weird words. Fourier time is the amount of time it takes heat to reach one location from another. E.g. i warm a bar one end how long does it take to reach the other. Applied to steady state theory if i apply a pulse (like a hot bit of processor warming up) whats the mean time for me to see an output of that pulse at the end. Its going to be governed by the bodies thermal mass and its conductivity. In the case of a solid bit of copper its a diffusive process wasily workable (if like abit of maths and simplifying assumptions) from basic laws. In heat pipes you are dealing with a much slower mass transfer process. Its introducing dynamic process into the mix (just to make the question harder). I'm unsure that it is really the case as the heat is going to quite rapidly diffuse. A little thought experiment proves that it is quite a weird problem that isn't really going to come up alot.

12-22-2007, 02:50 AM	#12
wildfrogman1 Cooling Neophyte Join Date: Mar 2006 Location: Usa Posts: 6	Re: Discussion of heatsink designs http://www.blairwing.com/images/ULTCo2/9.jpg I found this while looking around online. This seems relevant to the discussion. From what I've seen, and read certain heatpipe heatsink bases are overly thin. Sometimes to the point of base warping under high mounting pressure. So you need a thick enough base to keep from warping, yet focused cooling all across the die. The hottest die spots are not in the center of most cpu dies, but off on the side halfway between center and the corners. I got a birthday notice in my email, and remembered I had an account here Sooo I'll be more active.

09-11-2005, 06:16 PM	#2
jaydee Put up or Shut Up Join Date: Dec 2001 Location: Spokane WA Posts: 6,506	I would like to know why a cooler that keeps the CPU colder doesn't overclock better. What would cause that... That is the question that needs to be answered before designs are considered. Obviously the design would have to be around that solution.

09-11-2005, 06:31 PM	#3
Sin22 Cooling Neophyte Join Date: May 2003 Location: Singapore Posts: 66	Summarised what I was trying to say very succinctly jaydee. My initial thoughts had to do with heatspots/spikes on the die and how inidividual heatsinks coped with it. i.e. those with thicker bases managed to negate to some degree the effects of heatspots in comparison to those with thinner bps much like waterblocks (just as how the Cascade at some instances was shown to overclock better than comparitive blocks though temps were almost similar).

09-11-2005, 08:53 PM	#5
Sin22 Cooling Neophyte Join Date: May 2003 Location: Singapore Posts: 66	The way I saw it is that the Cascade/Storm focused the cooling across the die effectively. The same can be said for thicker BP waterblocks where the heat-distribution is more even like the MCW500x/MCW600x series. However. in a sense, you can't really focus the cooling of a heatsink onto the location of the temperature sensors and what not, much like we could with waterblocks. (Cutouts in the base would be the exception much like the Cooljag CJC6166A). That is what I felt was happening with the SLK900A & MCX462-V whereby the thicker baseplate of the MCX462-V allowed for better heat distribution and evening out of the heatspots. Now though, with the heatpipe/tower heatsinks how would this be different with the added medium of a heatpipe? To me, I've seen heatpipes as more like highways, just moving the heat from one point to another, allowing for greater cooling efficiency at the new location (and also expanding some heat through the actual movement). However, comparing the two heatsinks currently in question, the baseplates are almost identical in thickness, except the Hyper6 includes an aluminum heatsink above the heatpipe soldered area. Would this addition in turn act as an increased thickness baseplate? For the Tuniq Tower120, the thickness of the base between the IHS and the bottom of the heatpipes appears to only be 1.5mm. The bad thing in a sense with all this is that thermal resistance figures can be easily found and in sense double-checked versus other testbeds (though the die size is an issue), however, double-checking the overclocking effectiveness of a heatsink is quite based on anecdotal evidence.

09-11-2005, 11:50 PM	#6
Spot Cooling Neophyte Join Date: Nov 2003 Location: huh? Posts: 85	My guess is......... That explains the heat sink at the bottom of the hyper6. The extra metal below helps spread the heat load while the gas/liqiud in heat pipes takes their own sweet time pushing the heat through. Looks like the copper base of the Tuniq has a smaller threshold on the amount of heat it can take down there. Therefore, a simple heatsink attached to the copper base at the bottom may just help.

09-12-2005, 07:57 PM	#7
Sin22 Cooling Neophyte Join Date: May 2003 Location: Singapore Posts: 66	Sound plausible. So the only thing to do is to solder/attach a heatsink to the base in some fashion to see if overall overclocking increases.

11-11-2010, 02:10 PM	#13
koolguy Cooling Neophyte Join Date: Nov 2010 Location: Charlotte,nc Posts: 4	Re: Discussion of heatsink designs Hello everybody, I want to start playing around with oil cooling. Now, Submerging a MB in oil is totally nuts...But.. What I am thinking is something like a water cooling system but running oil in it. Does anybody know about any website to start reading about it? Thank you in advance.

Currently Active Users Viewing This Thread: 2 (0 members and 2 guests)